Quantifying Features of Musical Compositions

Musical compositions are complex objects with intricate patterns of pitch, rhythm, and timbre. In my undergraduate thesis, I explored numerical methods for finding identifying patterns in the works of famous composers. Music stored as MIDI files can be converted into a large array for analysis. Using this array, we explored patterns of pitch (melody), timing (rhythm), and chords (harmony). We sought to find common patterns in the works of famous composers with the ultimate goal of distinguishing between them automatically.

We drew on ideas from statistical mechanics in thinking about how emergent macroscopic observables can appear from averages of microscopic details. In this case, the microscopic details are pitches and timing of each note. The macroscopic details are the melodic, harmonic, and rhythmic patterns. These results were presented at the Midstates Conference for Undergraduate Research in Mathematics in 2007.

Collaborators:

• Eric Barth (Kalamazoo College)